On Jan 9, 2008 1:19 AM, Jonathan Ellis <jonathan@...> wrote:
> On Wed, 9 Jan 2008 00:14:54 -0500, "Jason Persampieri"
> <spyce-discussion@...> said:
> > It turns out that the spyce server was using the same 'spysession' temp
> > file
> > for the two runs. And seeing as how it was initially created as 'root',
> > the
> > measly 'apache' user couldn't access it. Is this by design? It seems
> > like
> > the server should start a new session file every time it's run, yes?
> Since the whole point of sessions is to make a stateless protocol more
> stateful, that doesn't necessarily follow at all in my mind. :)
> Think of it this way: if you were sharing a session db across N servers
> (spyce doesn't have a db backend yet, but hypothetically :) you wouldn't
> want to reset it if one server crashed and restarted for any N > 1. So
> really it would be confusing to blow it away if N = 1.
> (Yes, this is how all the other session systems I know of work too, it's
> not just a Spyce quirk.)
Um... yes... well.... that makes entirely too much sense :)
If I had known more about 'shelve' and the idea of a stateful server, I
would have found the error much faster. Ah well.
Perhaps there could simply be a more informative error message?